Overview
[Product Introduction] OpenViking is an open source context database designed specifically for AI Agents. It abandons the fragmented vector storage model of traditional RAG and innovatively adopts the "file system paradigm" to uniformly manage the Agent's memory, resources and skills. Through structured information organization, OpenViking is committed to solving problems such as context fragmentation, excessive token consumption, and unexplainable retrieval processes, helping developers build intelligent agents with long-term memory and self-evolution capabilities. [Product Features] File system paradigm: Map the context into a virtual file directory through the viking:// protocol, supporting standardized operations and unified management of memory and resources. Hierarchical context loading: Adopt L0(Abstract)/L1(Overview)/L2(Details) three-level loading strategy to obtain information on demand and significantly reduce Token costs. Directory recursive retrieval: Combining intent analysis, vector positioning and recursive drill-down, it provides a more accurate and global structured retrieval than single vector retrieval. Observable and self-evolving: supports full visual tracking of retrieval trajectories, and can automatically extract experience from conversational interactions to realize self-iteration of Agent memory. Project open source address: https://github.com/volcengine/OpenViking
Connected profiles
More Coding assistants products
Sources · 0Sources & provenance
English text is maintained separately from the original record. Consult the original source before citing a fact.
No source links are available for this profile.
Original source text
Names may use phonetic renderings when no English name is supplied. The original wording is retained below.
【产品介绍】OpenViking 是一款专为 AI Agent 设计的开源上下文数据库。它摒弃传统 RAG 的碎片化向量存储模式，创新性地采用“文件系统范式”统一管理 Agent 的记忆、资源与技能。通过结构化的信息组织，OpenViking 致力于解决上下文碎片化、Token 消耗过大及检索过程不可解释等难题，助力开发者构建具有长久记忆与自我演进能力的智能 Agent。【产品功能】文件系统范式：通过 viking：// 协议将上下文映射为虚拟文件目录，支持对记忆和资源进行标准化操作与统一管理。分层上下文加载：采用 L0（摘要）/L1（概览）/L2（详情）三级加载策略，按需获取信息，大幅降低 Token 成本。目录递归检索：结合意图分析、向量定位与递归下钻，提供比单一向量检索更精准、更具全局观的结构化检索。可观测与自演进：支持检索轨迹的全程可视化追踪，并能从会话交互中自动提取经验，实现 Agent 记忆的自我迭代。项目开源地址：https://github.com/volcengine/OpenViking